Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7f92805cb908>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7f928022b5f8>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.1.0
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    inputs_real = tf.placeholder(tf.float32,
                                 shape=(None, image_width, image_height, image_channels),
                                 name='input_real')

    inputs_z = tf.placeholder(tf.float32,
                              shape=(None, z_dim),
                              name='input_z')

    learning_rate = tf.placeholder(tf.float32, shape=(), name='learning_rate')

    return inputs_real, inputs_z, learning_rate


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).

def discriminator(images,
                  reuse=False,
                  alpha=0.2,
                  keep_prob=0.5):
    """
    Create the discriminator network
    :param images: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :param alpha

    :param keep_prob: if set will perform drop out on a few layers

    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """

    def leaky_relu(x):
        return tf.maximum(alpha * x, x)

    def drop_out(x):
        # drop out only if keep_prob is set
        return tf.nn.dropout(x, keep_prob=keep_prob) if keep_prob else x

    with tf.variable_scope('discriminator', reuse=reuse):
        # Input layer is 28x28x3
        x1 = tf.layers.conv2d(images, filters=64, kernel_size=5, strides=2, padding='same')
        relu1 = drop_out(leaky_relu(x1))

        #  we will reduce image dimensions by half and double the depth between each layers.

        # 14x14x64
        x2 = tf.layers.conv2d(relu1, filters=128, kernel_size=5, strides=2, padding='same')
        relu2 = drop_out(leaky_relu(tf.layers.batch_normalization(x2, training=True)))

        # 7x7x128
        x3 = tf.layers.conv2d(relu2, filters=256, kernel_size=5, strides=2, padding='same')
        relu3 = drop_out(leaky_relu(tf.layers.batch_normalization(x3, training=True)))

        # 4x4x256
        # Flatten it
        flat = tf.reshape(relu3, (-1, 4 * 4 * 256))
        logits = tf.layers.dense(flat, 1)
        out = tf.sigmoid(logits)

        return out, logits

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z,
              out_channel_dim,
              is_train=True,
              alpha=0.2,
              keep_prob=0.5):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :param alpha:
    :param keep_prob: if set will perform drop out on a few layers

    :return: The tensor output of the generator
    """

    def leaky_relu(x):
        return tf.maximum(alpha * x, x)

    def drop_out(x):
        # drop out only if keep_prob is set
        return tf.nn.dropout(x, keep_prob=keep_prob) if keep_prob else x

    with tf.variable_scope('generator', reuse=not is_train):
        # First fully connected layer

        #  to 2x2x512
        x1 = tf.layers.dense(z, 2 * 2 * 512)
        x1 = tf.reshape(x1, (-1, 2, 2, 512))
        x1 = drop_out(leaky_relu(tf.layers.batch_normalization(x1, training=is_train)))

        # to 7x7x256
        x2 = tf.layers.conv2d_transpose(x1, filters=256, kernel_size=5, strides=2, padding='valid')
        x2 = drop_out(leaky_relu(tf.layers.batch_normalization(x2, training=is_train)))

        # to 14x14x128
        x3 = tf.layers.conv2d_transpose(x2, filters=128, kernel_size=5, strides=2, padding='same')
        x3 = drop_out(leaky_relu(tf.layers.batch_normalization(x3, training=is_train)))

        # to 28x28 x out_channel_dim
        logits = tf.layers.conv2d_transpose(x3, filters=out_channel_dim, kernel_size=5, strides=2, padding='same')

        out = tf.tanh(logits)

        return out


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim, alpha=0.2):
    """
    Get the loss for the discriminator and generator
    :param alpha:
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    g_model = generator(input_z, out_channel_dim, alpha=alpha)
    d_model_real, d_logits_real = discriminator(input_real, alpha=alpha)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True, alpha=alpha)

    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, labels=tf.ones_like(d_model_real)))
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_model_fake)))
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_model_fake)))

    d_loss = d_loss_real + d_loss_fake

    return d_loss, g_loss



"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count,
          batch_size,
          z_dim,
          learning_rate,
          beta1, get_batches,
          data_shape,
          data_image_mode,
          print_every=10,
          show_every=100,
          show_images=25):
    """
    Train the GAN
    :param show_images:
    :param show_every:
    :param print_every:
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    image_width, image_height, image_channels = data_shape[1:]

    input_real, input_z, input_lr = model_inputs(image_width=image_width,
                                                 image_height=image_height,
                                                 image_channels=image_channels,
                                                 z_dim=z_dim)

    d_loss, g_loss = model_loss(input_real=input_real,
                                input_z=input_z,
                                out_channel_dim=image_channels)

    d_opt, g_opt = model_opt(d_loss=d_loss,
                             g_loss=g_loss,
                             learning_rate=learning_rate,
                             beta1=beta1)

    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        for epoch_i in range(epoch_count):
            for steps, batch_images in enumerate(get_batches(batch_size), start=1):
                # Sample random noise for G
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
                # Scale to  (-1, 1) from (-0.5,  0.5)
                batch_images = batch_images * 2.0

                # Run optimizers
                _ = sess.run(d_opt, feed_dict={input_real: batch_images, input_z: batch_z})
                _ = sess.run(g_opt, feed_dict={input_z: batch_z, input_real: batch_images})

                if steps % print_every == 0:
                    # At the end of each epoch, get the losses and print them out
                    train_loss_d = d_loss.eval({input_z: batch_z,
                                                input_real: batch_images,
                                                input_lr:learning_rate})

                    train_loss_g = g_loss.eval({input_z: batch_z,
                                                input_lr: learning_rate})

                    print("Epoch {}/{}...".format(epoch_i + 1, epoch_count),
                          "Batch: {}...".format(steps),
                          "Discriminator Loss: {:.4f}...".format(train_loss_d),
                          "Generator Loss: {:.4f}".format(train_loss_g))

                if steps % show_every == 0:
                    show_generator_output(sess,
                                          n_images=show_images,
                                          input_z=input_z,
                                          out_channel_dim=image_channels,
                                          image_mode=data_image_mode)

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 32
z_dim = 100
learning_rate = 0.002
beta1 = 0.5


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2... Batch: 10... Discriminator Loss: 1.5686... Generator Loss: 3.9427
Epoch 1/2... Batch: 20... Discriminator Loss: 0.5613... Generator Loss: 1.1231
Epoch 1/2... Batch: 30... Discriminator Loss: 1.1073... Generator Loss: 1.2429
Epoch 1/2... Batch: 40... Discriminator Loss: 0.0030... Generator Loss: 12.5638
Epoch 1/2... Batch: 50... Discriminator Loss: 0.4128... Generator Loss: 3.5659
Epoch 1/2... Batch: 60... Discriminator Loss: 0.3421... Generator Loss: 3.2980
Epoch 1/2... Batch: 70... Discriminator Loss: 3.7004... Generator Loss: 0.4750
Epoch 1/2... Batch: 80... Discriminator Loss: 0.2591... Generator Loss: 2.1935
Epoch 1/2... Batch: 90... Discriminator Loss: 0.8462... Generator Loss: 4.8435
Epoch 1/2... Batch: 100... Discriminator Loss: 3.5097... Generator Loss: 0.4305

Epoch 1/2... Batch: 110... Discriminator Loss: 0.7309... Generator Loss: 4.3317
Epoch 1/2... Batch: 120... Discriminator Loss: 0.7209... Generator Loss: 4.8077
Epoch 1/2... Batch: 130... Discriminator Loss: 0.2882... Generator Loss: 4.2812
Epoch 1/2... Batch: 140... Discriminator Loss: 1.0700... Generator Loss: 1.5384
Epoch 1/2... Batch: 150... Discriminator Loss: 0.4825... Generator Loss: 2.9277
Epoch 1/2... Batch: 160... Discriminator Loss: 0.5589... Generator Loss: 1.6926
Epoch 1/2... Batch: 170... Discriminator Loss: 0.2348... Generator Loss: 2.8008
Epoch 1/2... Batch: 180... Discriminator Loss: 0.1514... Generator Loss: 10.0028
Epoch 1/2... Batch: 190... Discriminator Loss: 1.0223... Generator Loss: 2.9017
Epoch 1/2... Batch: 200... Discriminator Loss: 0.2266... Generator Loss: 3.8449

Epoch 1/2... Batch: 210... Discriminator Loss: 0.5036... Generator Loss: 2.9825
Epoch 1/2... Batch: 220... Discriminator Loss: 0.1779... Generator Loss: 4.7665
Epoch 1/2... Batch: 230... Discriminator Loss: 0.4556... Generator Loss: 4.1712
Epoch 1/2... Batch: 240... Discriminator Loss: 0.4205... Generator Loss: 3.0610
Epoch 1/2... Batch: 250... Discriminator Loss: 0.5230... Generator Loss: 3.6248
Epoch 1/2... Batch: 260... Discriminator Loss: 0.9817... Generator Loss: 4.7945
Epoch 1/2... Batch: 270... Discriminator Loss: 0.7313... Generator Loss: 1.3564
Epoch 1/2... Batch: 280... Discriminator Loss: 1.0595... Generator Loss: 0.9504
Epoch 1/2... Batch: 290... Discriminator Loss: 0.8164... Generator Loss: 2.6687
Epoch 1/2... Batch: 300... Discriminator Loss: 1.1559... Generator Loss: 1.2207

Epoch 1/2... Batch: 310... Discriminator Loss: 1.3630... Generator Loss: 1.5863
Epoch 1/2... Batch: 320... Discriminator Loss: 1.0980... Generator Loss: 1.2481
Epoch 1/2... Batch: 330... Discriminator Loss: 1.4570... Generator Loss: 1.8253
Epoch 1/2... Batch: 340... Discriminator Loss: 0.6388... Generator Loss: 1.4903
Epoch 1/2... Batch: 350... Discriminator Loss: 0.4801... Generator Loss: 3.1235
Epoch 1/2... Batch: 360... Discriminator Loss: 0.5234... Generator Loss: 4.1280
Epoch 1/2... Batch: 370... Discriminator Loss: 0.8270... Generator Loss: 3.0112
Epoch 1/2... Batch: 380... Discriminator Loss: 0.6226... Generator Loss: 1.7706
Epoch 1/2... Batch: 390... Discriminator Loss: 0.9446... Generator Loss: 1.2538
Epoch 1/2... Batch: 400... Discriminator Loss: 0.3497... Generator Loss: 3.5909

Epoch 1/2... Batch: 410... Discriminator Loss: 1.1308... Generator Loss: 5.6508
Epoch 1/2... Batch: 420... Discriminator Loss: 1.0229... Generator Loss: 1.2357
Epoch 1/2... Batch: 430... Discriminator Loss: 0.4875... Generator Loss: 2.1298
Epoch 1/2... Batch: 440... Discriminator Loss: 0.7012... Generator Loss: 1.8685
Epoch 1/2... Batch: 450... Discriminator Loss: 0.7391... Generator Loss: 4.1623
Epoch 1/2... Batch: 460... Discriminator Loss: 0.8301... Generator Loss: 1.4701
Epoch 1/2... Batch: 470... Discriminator Loss: 3.0399... Generator Loss: 0.3260
Epoch 1/2... Batch: 480... Discriminator Loss: 0.6004... Generator Loss: 2.0481
Epoch 1/2... Batch: 490... Discriminator Loss: 0.8029... Generator Loss: 1.1399
Epoch 1/2... Batch: 500... Discriminator Loss: 1.2927... Generator Loss: 2.7467

Epoch 1/2... Batch: 510... Discriminator Loss: 0.8371... Generator Loss: 3.0027
Epoch 1/2... Batch: 520... Discriminator Loss: 1.2603... Generator Loss: 1.3541
Epoch 1/2... Batch: 530... Discriminator Loss: 1.0682... Generator Loss: 1.4278
Epoch 1/2... Batch: 540... Discriminator Loss: 1.5707... Generator Loss: 1.0245
Epoch 1/2... Batch: 550... Discriminator Loss: 1.3922... Generator Loss: 0.8106
Epoch 1/2... Batch: 560... Discriminator Loss: 1.6498... Generator Loss: 0.4125
Epoch 1/2... Batch: 570... Discriminator Loss: 0.8692... Generator Loss: 2.4542
Epoch 1/2... Batch: 580... Discriminator Loss: 1.4629... Generator Loss: 2.8090
Epoch 1/2... Batch: 590... Discriminator Loss: 0.7472... Generator Loss: 2.7586
Epoch 1/2... Batch: 600... Discriminator Loss: 0.9428... Generator Loss: 1.8886

Epoch 1/2... Batch: 610... Discriminator Loss: 1.4466... Generator Loss: 0.8832
Epoch 1/2... Batch: 620... Discriminator Loss: 0.8070... Generator Loss: 1.6084
Epoch 1/2... Batch: 630... Discriminator Loss: 0.9804... Generator Loss: 2.3018
Epoch 1/2... Batch: 640... Discriminator Loss: 1.5085... Generator Loss: 1.0850
Epoch 1/2... Batch: 650... Discriminator Loss: 0.6595... Generator Loss: 1.2601
Epoch 1/2... Batch: 660... Discriminator Loss: 0.8180... Generator Loss: 1.5602
Epoch 1/2... Batch: 670... Discriminator Loss: 0.8491... Generator Loss: 1.4696
Epoch 1/2... Batch: 680... Discriminator Loss: 0.9064... Generator Loss: 2.4520
Epoch 1/2... Batch: 690... Discriminator Loss: 1.9734... Generator Loss: 0.3415
Epoch 1/2... Batch: 700... Discriminator Loss: 0.8189... Generator Loss: 1.8229

Epoch 1/2... Batch: 710... Discriminator Loss: 1.3991... Generator Loss: 0.9330
Epoch 1/2... Batch: 720... Discriminator Loss: 1.1483... Generator Loss: 2.2496
Epoch 1/2... Batch: 730... Discriminator Loss: 0.9047... Generator Loss: 2.1137
Epoch 1/2... Batch: 740... Discriminator Loss: 1.7084... Generator Loss: 0.5845
Epoch 1/2... Batch: 750... Discriminator Loss: 0.6410... Generator Loss: 1.3672
Epoch 1/2... Batch: 760... Discriminator Loss: 1.2106... Generator Loss: 1.2463
Epoch 1/2... Batch: 770... Discriminator Loss: 1.3740... Generator Loss: 1.0298
Epoch 1/2... Batch: 780... Discriminator Loss: 0.9438... Generator Loss: 2.2788
Epoch 1/2... Batch: 790... Discriminator Loss: 0.8652... Generator Loss: 2.1200
Epoch 1/2... Batch: 800... Discriminator Loss: 1.3004... Generator Loss: 1.6083

Epoch 1/2... Batch: 810... Discriminator Loss: 1.4912... Generator Loss: 0.6076
Epoch 1/2... Batch: 820... Discriminator Loss: 1.2168... Generator Loss: 1.7589
Epoch 1/2... Batch: 830... Discriminator Loss: 1.3217... Generator Loss: 0.4984
Epoch 1/2... Batch: 840... Discriminator Loss: 1.0953... Generator Loss: 1.4613
Epoch 1/2... Batch: 850... Discriminator Loss: 0.7222... Generator Loss: 0.9032
Epoch 1/2... Batch: 860... Discriminator Loss: 1.2444... Generator Loss: 0.7989
Epoch 1/2... Batch: 870... Discriminator Loss: 1.1321... Generator Loss: 0.8336
Epoch 1/2... Batch: 880... Discriminator Loss: 0.7242... Generator Loss: 1.5626
Epoch 1/2... Batch: 890... Discriminator Loss: 1.4724... Generator Loss: 0.6403
Epoch 1/2... Batch: 900... Discriminator Loss: 1.5008... Generator Loss: 1.3100

Epoch 1/2... Batch: 910... Discriminator Loss: 0.9843... Generator Loss: 1.8299
Epoch 1/2... Batch: 920... Discriminator Loss: 0.6585... Generator Loss: 1.2499
Epoch 1/2... Batch: 930... Discriminator Loss: 0.5752... Generator Loss: 1.6265
Epoch 1/2... Batch: 940... Discriminator Loss: 1.1730... Generator Loss: 2.7518
Epoch 1/2... Batch: 950... Discriminator Loss: 0.7527... Generator Loss: 1.7046
Epoch 1/2... Batch: 960... Discriminator Loss: 1.0024... Generator Loss: 2.1475
Epoch 1/2... Batch: 970... Discriminator Loss: 0.7111... Generator Loss: 1.6558
Epoch 1/2... Batch: 980... Discriminator Loss: 0.7747... Generator Loss: 1.1940
Epoch 1/2... Batch: 990... Discriminator Loss: 1.1857... Generator Loss: 2.3919
Epoch 1/2... Batch: 1000... Discriminator Loss: 1.1969... Generator Loss: 1.0407

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 16
z_dim = 100
learning_rate = .0002
beta1 = 0.25


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/1... Batch: 10... Discriminator Loss: 0.8628... Generator Loss: 1.3877
Epoch 1/1... Batch: 20... Discriminator Loss: 1.0498... Generator Loss: 1.7087
Epoch 1/1... Batch: 30... Discriminator Loss: 0.5037... Generator Loss: 2.0441
Epoch 1/1... Batch: 40... Discriminator Loss: 0.5411... Generator Loss: 2.5756
Epoch 1/1... Batch: 50... Discriminator Loss: 0.5890... Generator Loss: 2.1507
Epoch 1/1... Batch: 60... Discriminator Loss: 1.0116... Generator Loss: 2.3047
Epoch 1/1... Batch: 70... Discriminator Loss: 0.5468... Generator Loss: 2.4253
Epoch 1/1... Batch: 80... Discriminator Loss: 0.6669... Generator Loss: 2.2748
Epoch 1/1... Batch: 90... Discriminator Loss: 0.3552... Generator Loss: 2.8114
Epoch 1/1... Batch: 100... Discriminator Loss: 0.4201... Generator Loss: 2.1953

Epoch 1/1... Batch: 110... Discriminator Loss: 0.7117... Generator Loss: 2.2258
Epoch 1/1... Batch: 120... Discriminator Loss: 0.5774... Generator Loss: 2.3652
Epoch 1/1... Batch: 130... Discriminator Loss: 0.5394... Generator Loss: 2.0630
Epoch 1/1... Batch: 140... Discriminator Loss: 0.2608... Generator Loss: 2.3599
Epoch 1/1... Batch: 150... Discriminator Loss: 0.4019... Generator Loss: 2.2865
Epoch 1/1... Batch: 160... Discriminator Loss: 0.2797... Generator Loss: 3.3508
Epoch 1/1... Batch: 170... Discriminator Loss: 0.3366... Generator Loss: 3.2063
Epoch 1/1... Batch: 180... Discriminator Loss: 0.2616... Generator Loss: 2.5939
Epoch 1/1... Batch: 190... Discriminator Loss: 0.2833... Generator Loss: 2.5478
Epoch 1/1... Batch: 200... Discriminator Loss: 0.3217... Generator Loss: 2.9957

Epoch 1/1... Batch: 210... Discriminator Loss: 0.5702... Generator Loss: 3.8612
Epoch 1/1... Batch: 220... Discriminator Loss: 0.2719... Generator Loss: 4.1219
Epoch 1/1... Batch: 230... Discriminator Loss: 0.2360... Generator Loss: 3.3702
Epoch 1/1... Batch: 240... Discriminator Loss: 0.1239... Generator Loss: 3.7706
Epoch 1/1... Batch: 250... Discriminator Loss: 0.1394... Generator Loss: 2.6149
Epoch 1/1... Batch: 260... Discriminator Loss: 0.5249... Generator Loss: 2.3340
Epoch 1/1... Batch: 270... Discriminator Loss: 0.2444... Generator Loss: 3.6119
Epoch 1/1... Batch: 280... Discriminator Loss: 0.1260... Generator Loss: 2.9200
Epoch 1/1... Batch: 290... Discriminator Loss: 0.1262... Generator Loss: 3.9576
Epoch 1/1... Batch: 300... Discriminator Loss: 0.0734... Generator Loss: 3.5969

Epoch 1/1... Batch: 310... Discriminator Loss: 0.0942... Generator Loss: 3.9767
Epoch 1/1... Batch: 320... Discriminator Loss: 0.1810... Generator Loss: 2.5048
Epoch 1/1... Batch: 330... Discriminator Loss: 0.1633... Generator Loss: 3.9236
Epoch 1/1... Batch: 340... Discriminator Loss: 0.1913... Generator Loss: 2.8811
Epoch 1/1... Batch: 350... Discriminator Loss: 0.3571... Generator Loss: 3.1510
Epoch 1/1... Batch: 360... Discriminator Loss: 0.1336... Generator Loss: 4.5183
Epoch 1/1... Batch: 370... Discriminator Loss: 0.2318... Generator Loss: 3.9684
Epoch 1/1... Batch: 380... Discriminator Loss: 0.4919... Generator Loss: 4.3425
Epoch 1/1... Batch: 390... Discriminator Loss: 0.2604... Generator Loss: 2.5356
Epoch 1/1... Batch: 400... Discriminator Loss: 0.1776... Generator Loss: 4.4420

Epoch 1/1... Batch: 410... Discriminator Loss: 0.3397... Generator Loss: 4.2256
Epoch 1/1... Batch: 420... Discriminator Loss: 0.0518... Generator Loss: 3.8151
Epoch 1/1... Batch: 430... Discriminator Loss: 0.2132... Generator Loss: 4.1715
Epoch 1/1... Batch: 440... Discriminator Loss: 0.0443... Generator Loss: 3.7316
Epoch 1/1... Batch: 450... Discriminator Loss: 0.2348... Generator Loss: 3.4023
Epoch 1/1... Batch: 460... Discriminator Loss: 0.2831... Generator Loss: 3.1856
Epoch 1/1... Batch: 470... Discriminator Loss: 0.1657... Generator Loss: 4.4838
Epoch 1/1... Batch: 480... Discriminator Loss: 0.2624... Generator Loss: 4.6283
Epoch 1/1... Batch: 490... Discriminator Loss: 0.1266... Generator Loss: 3.8905
Epoch 1/1... Batch: 500... Discriminator Loss: 0.1725... Generator Loss: 4.7511

Epoch 1/1... Batch: 510... Discriminator Loss: 0.1681... Generator Loss: 5.7860
Epoch 1/1... Batch: 520... Discriminator Loss: 0.1139... Generator Loss: 2.9901
Epoch 1/1... Batch: 530... Discriminator Loss: 0.1354... Generator Loss: 4.2465
Epoch 1/1... Batch: 540... Discriminator Loss: 0.1113... Generator Loss: 4.9725
Epoch 1/1... Batch: 550... Discriminator Loss: 0.2107... Generator Loss: 3.7930
Epoch 1/1... Batch: 560... Discriminator Loss: 0.3239... Generator Loss: 4.5969
Epoch 1/1... Batch: 570... Discriminator Loss: 0.0597... Generator Loss: 5.4247
Epoch 1/1... Batch: 580... Discriminator Loss: 0.0953... Generator Loss: 4.3009
Epoch 1/1... Batch: 590... Discriminator Loss: 0.0131... Generator Loss: 6.6616
Epoch 1/1... Batch: 600... Discriminator Loss: 0.1401... Generator Loss: 3.9292

Epoch 1/1... Batch: 610... Discriminator Loss: 0.1598... Generator Loss: 3.4371
Epoch 1/1... Batch: 620... Discriminator Loss: 0.0340... Generator Loss: 5.8728
Epoch 1/1... Batch: 630... Discriminator Loss: 0.0426... Generator Loss: 5.8304
Epoch 1/1... Batch: 640... Discriminator Loss: 0.4831... Generator Loss: 5.8424
Epoch 1/1... Batch: 650... Discriminator Loss: 0.1614... Generator Loss: 5.8071
Epoch 1/1... Batch: 660... Discriminator Loss: 0.3108... Generator Loss: 4.5399
Epoch 1/1... Batch: 670... Discriminator Loss: 0.2482... Generator Loss: 3.6268
Epoch 1/1... Batch: 680... Discriminator Loss: 0.0416... Generator Loss: 4.9332
Epoch 1/1... Batch: 690... Discriminator Loss: 0.1766... Generator Loss: 5.9914
Epoch 1/1... Batch: 700... Discriminator Loss: 0.0263... Generator Loss: 4.6600

Epoch 1/1... Batch: 710... Discriminator Loss: 0.0157... Generator Loss: 6.1098
Epoch 1/1... Batch: 720... Discriminator Loss: 0.0151... Generator Loss: 6.4328
Epoch 1/1... Batch: 730... Discriminator Loss: 0.0079... Generator Loss: 6.5086
Epoch 1/1... Batch: 740... Discriminator Loss: 0.0494... Generator Loss: 5.0655
Epoch 1/1... Batch: 750... Discriminator Loss: 0.0653... Generator Loss: 4.2834
Epoch 1/1... Batch: 760... Discriminator Loss: 0.0464... Generator Loss: 4.8907
Epoch 1/1... Batch: 770... Discriminator Loss: 0.1264... Generator Loss: 4.9625
Epoch 1/1... Batch: 780... Discriminator Loss: 0.0713... Generator Loss: 3.8713
Epoch 1/1... Batch: 790... Discriminator Loss: 0.1240... Generator Loss: 5.1432
Epoch 1/1... Batch: 800... Discriminator Loss: 0.0154... Generator Loss: 6.4535

Epoch 1/1... Batch: 810... Discriminator Loss: 0.1160... Generator Loss: 6.2832
Epoch 1/1... Batch: 820... Discriminator Loss: 0.0408... Generator Loss: 9.3546
Epoch 1/1... Batch: 830... Discriminator Loss: 2.4493... Generator Loss: 14.8173
Epoch 1/1... Batch: 840... Discriminator Loss: 0.0505... Generator Loss: 5.0607
Epoch 1/1... Batch: 850... Discriminator Loss: 0.3540... Generator Loss: 11.0274
Epoch 1/1... Batch: 860... Discriminator Loss: 0.1571... Generator Loss: 3.9683
Epoch 1/1... Batch: 870... Discriminator Loss: 0.2006... Generator Loss: 4.5341
Epoch 1/1... Batch: 880... Discriminator Loss: 0.7746... Generator Loss: 9.6606
Epoch 1/1... Batch: 890... Discriminator Loss: 0.0984... Generator Loss: 6.1588
Epoch 1/1... Batch: 900... Discriminator Loss: 0.2786... Generator Loss: 3.6609

Epoch 1/1... Batch: 910... Discriminator Loss: 0.1712... Generator Loss: 4.4669
Epoch 1/1... Batch: 920... Discriminator Loss: 0.1392... Generator Loss: 8.2416
Epoch 1/1... Batch: 930... Discriminator Loss: 0.0212... Generator Loss: 5.2247
Epoch 1/1... Batch: 940... Discriminator Loss: 0.2558... Generator Loss: 7.5405
Epoch 1/1... Batch: 950... Discriminator Loss: 0.0992... Generator Loss: 4.7138
Epoch 1/1... Batch: 960... Discriminator Loss: 0.1527... Generator Loss: 3.3318
Epoch 1/1... Batch: 970... Discriminator Loss: 0.0858... Generator Loss: 4.4923
Epoch 1/1... Batch: 980... Discriminator Loss: 0.1980... Generator Loss: 3.5804
Epoch 1/1... Batch: 990... Discriminator Loss: 0.1236... Generator Loss: 5.4683
Epoch 1/1... Batch: 1000... Discriminator Loss: 0.4064... Generator Loss: 3.0178

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.